Papers
8
Total Citations
282
H-Index
6
About
Nutan Chen is a leading researcher in robot learning, reinforcement learning, and generative models, with a focus on enabling robots to learn complex, high-dimensional sensory skills. Her seminal work on stable reinforcement learning with autoencoders for tactile and visual data (142 citations) pioneered methods for robots to learn directly from high-dimensional feedback through trial and error, a critical step toward real-world dexterous manipulation. She also advanced movement representation by integrating dynamic movement primitives with variational autoencoders (57 citations), allowing efficient generalization of complex robotic movements. Chen has made notable contributions to evaluating deep generative models, proposing new metrics for VAEs and GANs (36 citations), and to active learning for robotics, developing methods that detect knowledge gaps during skill acquisition. Her research on estimating fingertip forces from images and fingernail deformations (14+ citations) bridges computer vision and biomechanics, with applications in prosthetic control and human-robot interaction. More recently, she is exploring language-informed multi-task visual world models for scalable robot learning. Chen’s work consistently addresses the challenge of making robots learn efficiently from high-dimensional, multimodal data.
Research Focus
Key Achievements
Top Papers
- 1Stable reinforcement learning with autoencoders for tactile and visual data142 citations · 2016
- 2
- 3Metrics for Deep Generative Models36 citations · 2017
- 4
- 5Active Learning based on Data Uncertainty and Model Sensitivity14 citations · 2018
- 6
- 7System architecture for an interactive patrolling humanoid robot4 citations · 2012
- 8LIMT: Language-Informed Multi-Task Visual World Models1 citations · 2025